PERSONALIZED TRAITS MONITORING USING A NEURAL NETWORK BASED ON OSCILLOMETRIC MEASUREMENTS

Abstract

Oscillometric method uses a pressure sensor instead of a stethoscope to record the pressure oscillations within the cuff. Oscillations observed from the pressure sensor show personalized traits of systolic pressure, which indicates the maximum amount of work the heart has to perform per stroke in order to move blood through the arteries. For a given person, the oscillation waveforms extracted from the cuff pressure have an oscillation pattern that varies in size over time. Thus, to extract uniform features from a given person, we normalize the variability of the corresponding oscillation patterns for different frequencies, and evaluate systolic blood pressure using a feedforward neural network based on the extracted features. The validity of measuring systolic blood pressure with a neural network was compared with the average values of systolic pressure obtained by two nurses using the auscultatory method. The recognition performance was found to be 98.2% for ±20 mmHg, 93.5% for ±15 mmHg, and 82.3% for ±10 mmHg based on the difference between the systolic pressures measured by the auscultation method and the proposed method with a neural network. The collected database for our experiment includes 85 participants with ages ranging from 10–80 years. In this paper, we present a novel personalized traits monitoring system that can monitor personalized traits of systolic blood pressure. This study can serve as a foundation for the diagnosis and management of isolated systolic hypertension.